data footprint = the trail of data your apps keep about you
In 3.1.1 you saw a feed turn your actions into a profile. Where did the actions come from? Every app you use keeps a record of some of what happens in it. The pile of records is your data footprint, and each app has its own piece of it.
Three ways an app gets data
A common way to sort personal data is by how the app got it. The line between these matters, because only one of them is something you told the app.
Given, observed, inferred
Given.You provided it on purpose: your email, a photo you post, a comment you write.
Observed.The app recorded it from what you do or from your device: when you opened it, how long you watched, where your phone was, what model it is.
Inferred.The app concluded it from the rest: "interested in hiking," "probably 15 to 17." These are guesses, and they are what your profile in 3.1.1 was made of.
Structured and unstructured data
Data also comes in two shapes. Structured data fits neatly into labeled fields, like a spreadsheet: a birthday, a time stamp, a number of seconds. Unstructured data is free-form: photos, videos, voice recordings, and the text of a comment. A computer can sort and add up structured data easily. Unstructured data is harder, and this is one place AI is used. A classifier looks at your photo and adds the tag #dog. A model reads your comment and labels its mood. The messy data becomes structured labels, which is exactly how tags ended up in your 3.1.1 profile.
⚠️Watch out: collecting data isn't the same as AI. Most collection is ordinary logging: the app writes down that something happened. AI enters when that data is used to infer, sort or predict. "The AI is watching me" skips the step that matters, which is what gets recorded and what gets concluded from it.
ExampleOne post, three kinds of data
Rio, an invented user, posts a photo of a dog on a trail to an invented app at 7:42 AM, with the caption "Saturday hike!" Sort what the app ends up with.
Step 1 — Given
The photo and the caption "Saturday hike!" Rio chose to share both. Both are unstructured: a picture and free text.
Try it
Sort the data · invented app, invented user
For each item pick how the app got it (given, observed, or inferred) and what shape it is (structured or unstructured).
A photo you posted
The time you opened the app: 7:42 AM
Your phone's location while the app was open
That you scrolled past a video in under 2 seconds
A profile label saying you're interested in hiking
A caption the app wrote for your photo: "a dog on a trail"
A prediction that you'll tap an ad for running shoes
The email address you typed to sign up
Correct so far: 0 / 8 · answers tried: 0
💡Your side of it: practice on the invented examples here. Don't look at classmates' phones, and don't paste real account details, screenshots of your data pages, or other people's information into any tool.
Principles of AI · 3.1.2SkillConcept
Reading a Privacy Label
privacy label = a short summary of what an app collects, why, and who else gets it
You don't have to guess what an app collects. Many app stores show a short privacy or data-safety summary on each app's page, and many apps have a privacy section in their settings. The names and layouts differ, so the labels in this course are invented, but the questions are the same.
Three questions for any label
What?Which categories of data does it collect?
Why?For which purpose: to make the app work, to run your account, to take payments, for analytics, or for advertising?
Who else?Is it shared with outside companies, and which kind?
The necessity test
Then ask one more thing about each row: could the app still do its job without it? If not, the data is needed. Directions need your location. If it could, the data is not needed for the job, and that is a good place to ask whether the trade is worth it to you.
ExampleBrightBeam, an invented flashlight app
BrightBeam's job is to turn on your phone's light. Read its label.
Step 1 — What?
It collects 3 categories: device & account ids, precise location, and your contacts list.
Now try it yourself. For each row, decide whether the data is needed for the app's job. Notice that sharing isn't automatically a problem: RouteBuddy's payment processor needs your purchase to take your payment.
Label reader · invented apps and labels
RouteBuddy · maps and rides app
Its job: get you from A to B
Precise locationApp function
Not shared with third parties
Needed for the job?
Contact infoAccount
Not shared with third parties
Needed for the job?
PurchasesPayments
Shared with: payment processor
Needed for the job?
Usage (taps, watch time)Analytics + Advertising
Shared with: ad partners
Needed for the job?
Collects 4 · shared with third parties 2 · used for advertising 1
🔑"Not needed for the job" is not the same as "bad." Analytics can help a developer fix problems. The label just shows you where the app wants more than it needs, so you can decide.
⚠️A label is the company's own statement. It can be out of date, and it doesn't prove what actually happens to the data. Treat it as a good starting point, not a guarantee. Permissions, like the pop-up asking to use your camera or location, are a separate control: they decide what an app can reach on your phone, while the label says what the company says it collects and why. Use both.
Principles of AI · 3.1.2ConceptSkill
Your Footprint Adds Up
aggregation = joining small pieces of data into one bigger picture
Each app holds only a slice of your footprint, and one slice may seem harmless. Pieces can be joined, though. Apps carry IDs for your account and your device, and phones often have an advertising ID that ad companies use to recognize the same device across apps. When two companies receive the same ID, they can match their pieces. The result can say more than any one piece did.
Whether two particular companies actually match their data depends on who gets what, so treat the lab below as showing what is possible, not what is happening. And like every inference in this unit, a combined picture is still a guess.
ExampleRio's three apps
Rio, an invented user, uses a short video app, a maps & rides app, and a fitness app. The lab's made-up "typical" labels for those kinds are on the page below.
Step 1 — Add up the categories
Between them the three kinds collect 8 of 10 categories.
Run your own audit
Audit lab · app kinds only · nothing here is saved
Tick the kinds of apps you use. Don't type app names or account details. The kinds below are made up and simplified, and real apps differ.
How many of your app kinds collect each category
Contact info
0
Device & account IDs
0
Precise location
0
Photos & videos
0
Your contacts list
0
Usage (taps, watch time)
0
Search history
0
Purchases
0
Health & fitness
0
Crash logs
0
What could combining them say? (guesses, not facts)
None of the combinations in this lab come up yet. Try ticking more kinds.
🧾 My Audit Card
App kinds: 0
Categories collected: 0 of 10
Most common: —
Combinations possible: 0 of 5
Share only these counts with a partner. These are what's possible if the pieces were linked, not proof that anyone is doing it.
What you can do
No setting makes you invisible, and some collection is needed for an app to work. But you have more choices than "accept everything" or "don't use apps." What's available changes from app to app, so look at the settings of the ones you use.
Check before you install. Read the label or settings summary. Ask what, why, and who else.
Allow only what the job needs. When an app asks for a permission, run the necessity test. If your device offers choices like "only while using" or "approximate location," pick the smallest one that still works.
Look for ad and data controls. Many apps and phones let you limit personalized ads or reset your advertising ID.
Post less of what you give. Given data is the part you control most directly.
Ask for deletion if it's offered. Deleting an app usually stops new collection, but data already collected may be kept. Some companies let you request a copy or deletion of your data, and what you can ask for depends on the company and where you live.
Use fewer apps for the same job. A smaller set of apps is a smaller footprint.
💡Your side of it: some of the data on your phone belongs to other people. When you let an app read your contacts list, you share your friends' names and numbers too, and they didn't agree. Ask before you upload someone else's photos or details.
🔑In 3.1.3 you'll audit one platform in depth and use what you find to ask a company for a change. The label-reading skill from this page is what you'll use.